The application of evolutionary computation to the diagnosis and monitoring of Parkinson's disease

Deep sequencing technologies are routinely employed to probe protein-DNA interactions and other epigenetic modifications such as DNA methylation. These technologies return complex results in the form of binding profiles which can be extended over several kb and present characteristic spatial patterns. Such patterns appear to be reproducible among replicates, yet how they can be incorporated in a formal statistical analysis of the data is non-trivial.

In this talk I will introduce MMDiff, a kernel-based statistical methodology to test for changes in spatial patterns in ChIP-Seq peaks. I will show how MMDiff analysis helped us formulate mechanistic hypotheses on the role of transcription factor proteins in the establishment of histone marks, which we later showed to be consistent with a large-scale analysis of the ENCODE data sets. Finally, if there is time, I will introduce M3D, an extension of MMDiff to DNA methylation profiles obtained with bisulfite-sequencing experiments.